Abstract Great competitive results of Slovenian ski jumpers in world cup and continental competitions have sparkled a lot of interest for active participation in this attractive sport. In junior levels, national competitions with considerably more than 100 jumpers are becoming the norm. However, due to lack of technologic aids for distance measurement, such competitions can last over half a day. Only... Great competitive results of Slovenian ski jumpers in world cup and continental competitions have sparkled a lot of interest for active participation in this attractive sport. In junior levels, national competitions with considerably more than 100 jumpers are becoming the norm. However, due to lack of technologic aids for distance measurement, such competitions can last over half a day. Only at the top-level competitions (world cup, continental cup) expensive and logistically demanding commercial video distance measuring tools are used for this purpose. In a previous project we developed a video distance measuring system from low-cost commercial components, which was not suitable for real-time usage due to technological limitations, but worked great for offline measurement. We analyze the results of offline measurements for several competitions and show that measurement errors are often unacceptably high. This serves as a motivation for an ongoing project, where video measurement is performed in real time and supported by advanced computer vision and deep learning methods. Read More Read Less

Abstract researched. Most commonly classification considers all activities to be ‘equal’ (we will use term at classification). However, intuition suggests better results could be achieved using a hierarchical approach for classification. In this paper we compare three different approaches to classify activities: (i) Flat classification - classes are equal and we build one model to classify all of them; (ii) Multi-model hierarchical classification -... researched. Most commonly classification considers all activities to be ‘equal’ (we will use term at classification). However, intuition suggests better results could be achieved using a hierarchical approach for classification. In this paper we compare three different approaches to classify activities: (i) Flat classification - classes are equal and we build one model to classify all of them; (ii) Multi-model hierarchical classification - classes are arranged in trees, we build different models to classify activities on different levels. We apply two different approaches; (iii) Hierarchical classification using CLUS software1. Read More Read Less

Abstract Fault Detection and Isolation (FDI) procedure increase assurance on quality, reliability and safety in industrial systems. A suitable installed sensors in an industrial process is a necessary condition for fault diagnosis. Sensor placement for diagnosis purposes consists to study which process variables have to be measured to satisfy diagnosis specifications. Analytical redundancy relations (ARRs) are used frequently in the area... Fault Detection and Isolation (FDI) procedure increase assurance on quality, reliability and safety in industrial systems. A suitable installed sensors in an industrial process is a necessary condition for fault diagnosis. Sensor placement for diagnosis purposes consists to study which process variables have to be measured to satisfy diagnosis specifications. Analytical redundancy relations (ARRs) are used frequently in the area of diagnosis as well as optimizing, analyzing, and validating of sensors of the system. This paper presents the optimal sensor placement approach based on structural analysis methods using tripartite graph approach. The proposed approach allows to study which sensors are required to be installed in a process in order to improve certain fault diagnosis specifications; and it includes 2 phases: (i) development of systematic and efficient approaches for the generation of ARRs set allowing to generate “influence matrix”, (ii) then the multi-criteria optimization is applied by selecting robust sensor placements (Pareto Optimal Solutions) leading to a sensor placement algorithm. The proposed method has been validated on a robot dynamic model taken as a benchmark where the benefits of the method are clearly shown. Read More Read Less